The placenta is the first organ to develop, vascularised chorionic villi being already evident during the third week of gestation. Indeed, rapid early development is a prerequisite for normal progression of the embryo. From these early stages, spatial segregation gives rise to two populations of placental villi, one that floats freely in the intervillous space, and a second population (anchoring villi) that acts to stabilise the mechanical integrity of the placental-maternal interface [1–4]. The surface of both types of structure is covered by trophoblast. The floating villi exhibit a bilayered epithelium consisting of cytotrophoblast with overlying syncytium, while the anchoring sites lack a continuous syncytium, featuring instead villous tips with multilayered columns of largely mononuclear cytotrophoblast [1, 5]. Villous cytotrophoblasts act as stem cells for both pathways. The syncytium functions as a transporting epithelium, while extravillous cytotrophoblasts in the distal columns provide physical anchorage of the villous tree, failure of which is likely to lead to miscarriage. The distal columns also act as feeder sites for a large population of migratory cytotrophoblasts that infiltrate the decidua and inner myometrium, where they colonise and transform the spiral arteries to allow increased blood flow to the intervillous space [6]. Impaired transformation of the myometrial arterial segments is associated with pregnancy pathologies including pre-eclampsia [7, 8]. Such considerations underline the importance of understanding the mechanisms that control trophoblast differentiation and entry to the migratory or syncytial pathways.

Investigations of events that occur during formation of anchoring sites and the signals that lead to extravillous trophoblast differentiation require experimental systems that replicate as faithfully as possible placental development in vivo. We have used an explant model [5, 9–11] in which stem villous cytotrophoblast at the tips of mesenchymal villi is stimulated by attachment to an extracellular matrix (ECM) to undergo a proliferative burst. This is followed by cell reorganisation into a radial outgrowth that resembles the cytotrophoblast columns and shell seen in vivo, both in terms of morphology and the progressive expression of molecular markers of the extravillous pathway [9, 10]. The process depends on interaction between fibronectin deposited by the cytotrophoblasts and its receptor, integrin α5β1, at the cell surface [10]. However, cell detachment from the distal column, migration and infiltration of the surrounding ECM as seen in vivo are not observed in vitro.

These findings suggest that whilst contact with a permissive ECM triggers the development of cytotrophoblast columns that function effectively in tissue anchorage, it is insufficient to stimulate the acquisition of a migratory phenotype. Thus, cell migration apparently requires a further stimulus. Using a placenta-decidua coculture system we have tested the possibility that this might arise from maternal decidual cells, but the results were remarkably similar to those seen in explants on cell-free ECM, suggesting that decidua may not be the source of a paracrine stimulus to trophoblast migration [9, 12]. In the light of these findings we have now investigated another potential source of paracrine effectors, namely the mesenchymal cell population that lies beneath the villous cytotrophoblast. At the base of cytotrophoblast columns, mesenchymal cells extend processes that approach and even make direct contact with the trophoblast basement membrane [4]. In the current investigation we show that first trimester fibroblasts produce signals that activate cytotrophoblasts to migrate from the periphery of nascent columns, and that one component of the paracrine repertoire is insulin-like growth factor (IGF).

In previous work, we showed that contact between the tips of first trimester mesenchymal villi and a permissive ECM stimulates the development of anchoring columns containing cytotrophoblasts that initially proliferate, and then differentiate in a pattern that resembles closely that seen in columns in vivo [9, 10]. However, the cells did not advance to a migratory pattern of behaviour, suggesting that a further stimulus is required. The present results identify placental mesenchyme as one source of this signal (Fig. 8). Developmentally, this is appropriate, as the investment of primary villi by extraembryonic mesenchyme occurs within the first few days after implantation, a time at which trophoblast migration is beginning [4]. The need for mesenchymal stimuli to the trophoblastic epithelium during placental development has also been identified in the mouse, where the gcm1 gene product, which is needed for formation of the labyrinth, is stimulated in basal chorionic trophoblast by contact with the allantoic mesoderm [17].

The presence of IGFs at the maternal-fetal interface has been documented both in humans and a variety of animal species [18–20]. The importance of the IGF system to pregnancy outcome has been clearly demonstrated by the severe fetal growth restriction that occurs in mice lacking either IGF-I or IGF-II [19]. There is evidence that fetal growth is dependent on IGF in humans [18]. We have now identified IGF-I as an important component in placental mesenchymal-epithelial signalling. A role for the IGF system in promoting cytotrophoblast migration was suggested by the discovery that the extravillous cytotrophoblast of first trimester express mRNA encoding IGF-II [20, 21], together with the demonstration that IGF-II stimulates trans-filter migration in cultured cytotrophoblast [14, 15]. From these findings arose the idea of an autocrine loop in which trophoblast might enhance its own migration.

We have used an explant model in which de novo development of columns occurs specifically at the distal tips of first trimester mesenchymal villi [10, 11]. Columns retain normal polarity, with a stromal compartment beneath the villous basement membrane. In this model IGF-II and activin AA [13] stimulate cytotrophoblast migration, providing further evidence in favour of autocrine effects. Addition of IGF-1 however produces a striking streaming pattern of cytotrophoblast migration. The production of IGF-I in vitro by fibroblasts is consistent with low but significant levels of IGF-I mRNA in this location in vivo [19] as well as immunoreactivity indicating that the factor is present. IGF-I appears not to be produced by trophoblasts. However cytotrophoblast in columns expresses the receptors IGFR1 and IR as well as their intracellular target IRS-1. A fraction of placental IGFR1 heterodimerises with IR [22]. IGFs can act through the insulin receptor [23].

We therefore postulate the existence of a paracrine mechanism that stimulates trophoblast migration (Fig. 8). Further work will be required to investigate the part played by the IGF/insulin receptor family in trophoblast migration, including other intracellular targets and the phenotypic alterations that confer increased motility. There are suggestions that other receptor species including the IGFR2/M6P receptor may be capable of signal transduction and effects on trophoblast motility [15, 24].

In this model, as trophoblast migrates more deeply into the decidua and myometrium, it progresses farther away from the villous mesenchyme where IGF-I originates, so the magnitude of the signal must clearly diminish. This suggests the possibility of a self-limiting mechanism for migration, in which local inhibitory stimuli eventually outweigh the stimulatory signal. The reason for the presence of both autocrine (IGF-II) and paracrine (IGF-I) stimuli is as yet unclear, though IGF-I produces a more notable streaming phenotype. An alternative interpretation is that the paracrine trigger might act simply as a switch to a migratory pattern of behaviour. The two IGFs differ in their effects on proliferation and differentiation of rat trophoblast [25]. The role of IGF binding proteins and in particular IGFBP-1, which is a major secretory product of decidual cells [18, 20], in regulation of the IGF system at the maternofetal interface remains to be resolved [14, 18, 26].

Extravillous cytotrophoblast differentiation requires at least two distinct steps (Fig. 8). In the first step, a column forms as a result of attachment to ECM. In the second step, IGF-I from placental mesenchymal cells stimulates progression to a fully migratory phenotype. This leads to a model in which, as the trophoblast migrates further away form its site of origin at the anchoring villus, through the decidua and into the myometrium, the migratory stimulus becomes diluted. Thus a mechanism is suggested whereby trophoblast migration might cease in the inner myometrium. Escape from paracrine regulation is likely to be a feature of trophoblastic tumours. Whether the IGF system also plays a role in the normal colonisation and transformation of maternal spiral arteries, and whether it may be affected in the major pregnancy pathologies of pre-eclampsia, IUGR and spontaneous abortion, remains to be investigated.

Collagen type I (rat tail tendon) was obtained from Beckton Dickenson (Oxford, UK) and Matrigel from Stratech Scientific (Luton, UK). Dulbecco's Eagle's Medium (DMEM), Ham's F-12 and fetal calf serum (FCS) were from Gibco (Life Technologies, Paisley, UK). Human Activin AA was a kind gift from the National Hormone and Pituitary Program, NIDDK (Dr Parlow). Rabbit anti-mouse streptavidin-FITC and biotinylated rabbit anti-mouse-Ig were from Dako (High Wycombe, UK). Human recombinant IGF-I was obtained from Genentech (San Francisco, USA), recombinant human TGFβ1 from R&D Systems (Abingdon, UK). Monoclonal anti-cytokeratin 7 was from Dako (Ely, UK). Monoclonal antibody αIR3 to IGFR1 was from Calbiochem (Nottingham, UK). Monoclonal Ab-1 to IGFR1 was a kind gift from Prof K Siddall, University of Cambridge, UK. Monoclonal antibody 15C9 to IGF-I was as described previously [27].